BIND 10 experiments/vorner/receptionist, created. fdb6fd8e3a4e680d6739835bc48510d2d44eb866 [2776] Minor tweaks and fixes

[BIND 10 source code commits]

The branch, experiments/vorner/receptionist has been created
        at  fdb6fd8e3a4e680d6739835bc48510d2d44eb866 (commit)

- Log -----------------------------------------------------------------
commit fdb6fd8e3a4e680d6739835bc48510d2d44eb866
Author: Mukund Sivaraman <muks at isc.org>
Date:   Thu Jun 13 11:13:08 2013 +0200

    [2776] Minor tweaks and fixes

commit 8094aaa11df22d79f338539b338e07bd8f950117
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Tue Mar 5 10:07:42 2013 +0100

    [2776] Provide a proposal for the design
    
    As a result of the experiment, come to the conclusion the receptionist
    is easy enough to implement, flexible enough and fast enough to try and
    implement it.

commit da274764a6096182c8a64b5f66cfecaa715a470c
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Thu Feb 28 14:31:14 2013 +0100

    [2776] Tweak the receptionist

commit af52109c9d749dc249349dadcba64c07db18b9ff
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Thu Feb 28 13:08:08 2013 +0100

    [2776] Disable sync optimisation for UDP
    
    For the other test, we use TCP and that one doesn't have the
    optimisation. Disable it for UDP too so it is more fair.

commit 2450e94f926e81bed1af7488dd3c3d2fd58d5872
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Thu Feb 28 12:47:15 2013 +0100

    [2776] Forward replies

commit 9a5b9919090672e9c01282f6f75e91a66ec5439c
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Thu Feb 28 11:05:47 2013 +0100

    [2776] Don't parse the stack as DNS message
    
    Also add many debug outputs to help finding it out.

commit abe72e9dff228269f889379f63228001e807982c
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Wed Feb 27 13:43:10 2013 +0100

    [2776] Forward messages
    
    Basics of the receptionists. Forward the messages to auth. It doesn't
    seem to work properly, though. No idea if the error is on the auth or
    receptionist side.

commit b859ed9695b1b4fb1a8861672a05e37065bc1593
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Wed Feb 27 10:38:53 2013 +0100

    [2776] Let the TCP server accept batches of messages
    
    Each batch is preceded by 4-byte length of the batch. Then the messages
    are put one after another, with the usual 2-byte length of message.
    
    This should allow for less reads and less context switches in the
    following experiment.
    
    As this is experiment, no tests are updated or written. The old ones,
    obviously, don't work, as this is backwards-incompatible change. Hoping
    it just works, if not, it'll be debugged later when the other part is
    written.
    
    Also, the code style is probably not clean either, comments and docs are
    missing, etc.
    
    This will work for auth, but probably not for the resolver (or, it
    would, but not well, as it would block for one message before starting
    another). This is not an issue with auth server, as all the lookups are
    synchronous.

commit 6360d78376848609b4d19e8e8282dbc76d9c41e1
Author: Michal 'vorner' Vaner <michal.vaner at nic.cz>
Date:   Tue Feb 26 12:45:12 2013 +0100

    [2776] Some design considerations for sharing a socket
    
    Write some advantages and disadvantages of each method how to solve it.

-----------------------------------------------------------------------